Large scale implementation of Serendipity

[kmccann]

1) If I want to have 100,000 Serendipity blog sites (perhaps many more), what are the constraints? (assume many would be under one or a few domains, with URL's like www.blogs.com/kmccann

2) If I'm not mistaken, I can do this with a single database and use prefixed tables instead of creating a DB for each site? Is this correct? Would this be the best approach?

3) For a large-scale implementation, what advantages does Serendipity have over other blogging software? Are there others that would be better?

Keep in mind that we do not want to use a service, we want to install and run blogging software, create our own plugins, etc. I really like Serendipity for my personal blogging purposes and would like to be able to present the advantages of using it in a large-scale implementation.

Thanks,
Kevin

[garvinhicking]

Hi!

1) If I want to have 100,000 Serendipity blog sites (perhaps many more), what are the constraints? (assume many would be under one or a few domains, with URL's like www.blogs.com/kmccann

Just the usual PHP application issues - you might need to think about reverse proxy setups, using lighthttpd+php-cgi and RAM+RAM+RAM+RAM. And RAM won't hurt. As well as a good CPU and RAM.

You might also want to run a clustered setup where you only run a few thousand blogs on one machine to reduce the total load.

2) If I'm not mistaken, I can do this with a single database and use prefixed tables instead of creating a DB for each site? Is this correct? Would this be the best approach?

You could. But each s9y install has 20 tables, so you'd have 2.000.000 tables on thet single database. I think I've heard that mySQL doesn'T deal that well with so many tables in one single database.

Also, with all blogs on a single MySQL server you will have vast performance issues on that MySQL server. You should also best creat a replication clustered mysql server for this.

3) For a large-scale implementation, what advantages does Serendipity have over other blogging software? Are there others that would be better?

Serendipity is not really optimized for large-scale implementation, so there are some issues where one could work at for improving it (like better caching, full static pregeneration). On the other hand, no blog application I know if is optimized for this, yet.

MoveableType comes closest large-scale implementation, but with its perl approach is a larger pain to work with to circumvent typical regeneration problems.

I'd be really interested in such a large scale setup, and could maybe throw a few ideas at the people implementing it, or work together with your people in thinking of best code changes to suit your needs, so that the serendipity distribution could benefit frmo that.

Best regards,
Garvin

[kmccann]

Hi Garvin,

Thanks for your response. We currently have a clustered environment with load balancing for our busy web and email services. We use a Master/Slave MySQL setup for database storage. I don;t knwo much about the clustered mysql solution yet. We may start using Amazon Cloud clusters for newer services such as this blogging stuff.

You bring up some good points to keep in mind. We'll be evaluating a small number of blogging tools (candidates will be chosen based on various criteria). If Serendipity remains in the running, it would be great to have you involved. Could be a mutually beneficial learning experience. I have your email address from when I needed help with something else, and will follow up with a private message.

Thanks,
Kevin